Pricing
Transparent pricing, in kroner.
Priced in NOK, metered live, with hard quotas instead of surprise bills. All prices ex VAT and indicative — enterprise agreements are individually negotiated.
Managed models & Your models
Inference
- 20 NOK
- per 1M input tokens
- 60 NOK
- per 1M output tokens
- from 40 NOK
- per GPU-hour of endpoint time — billed per second, only while the endpoint is running
Applies to catalog models and to your own uploaded models alike. Per-project keys, rate limits and live NOK metering included.
GPU compute
Dedicated GPUs
- from 40 NOK
- per GPU-hour
- Per second
- billing granularity — stopped instances cost nothing
- Whole GPUs
- only — your GPU is never shared with another tenant
SSH access and curated base images included. Fine-tune on your data, then serve the result as a managed endpoint.
Enterprise
Air-gapped & on-prem
Custom deployments for environments where the uplink does not exist. Includes:
- Mirrored container registry
- Offline model transfer
- Compliance documentation (NSM, GDPR, DORA)
- Support SLAs
Pricing questions
- How granular is billing?
- GPU time is billed per second, from the moment a resource starts until you stop it. Stopped instances and paused endpoints cost nothing. Token usage is metered per request and aggregated to your project ledger in real time.
- What counts as a token?
- Tokens are counted by the tokenizer of the model serving your request — the same count returned in every API response's usage field. Input tokens cover your prompt and context; output tokens cover the generated completion.
- What happens when I hit my quota?
- Quotas are hard limits. When a project reaches its configured quota, requests are rejected until you raise the limit — there is no automatic overage and there are no surprise bills.
- Do prices include VAT?
- No. All prices are in Norwegian kroner (NOK), exclusive of VAT, and indicative. Enterprise agreements — including committed capacity and air-gapped deployments — are negotiated individually.